Inspiration
We had no idea what to do and then we realised that rabbits are actually a thing and exist on campus. We designed an SNN to hunt down rabbits whilst we chased them with a LiDAR and Ultrasonic sensor.
What it does
It takes in input via webcam continuously over time and feeds it into the input layer of an SNN which only receives detected differences between frames, reducing the overall amount of computation, and aims to detect objects similar to how a CNN does.
How we built it
We built it using PyCharm, numerous libraries (especially ml_genn), and a physical circuit which is no longer part of the original scope.
Challenges we ran into
Learning how to make an SNN. Learning how to perform sensor fusion (merging two different types of sensor data into one), and adapting an SNN for continuous input.
Accomplishments that we're proud of
Creating an SNN which receives continuous data over time in an efficient manner.
What we learned
We had learned how to make an SNN, how to pre-process time series data, and how to coordinate multiple sensors and reduce errors due to latent communication.
What's next for SNN Motion Detector
Potentially using discrete fourier transforms to convert the time series data into the frequency domain and seeing how the SNN responds to that.
Log in or sign up for Devpost to join the conversation.